Effect size for rectangular frequency table

Question

I have run several Chi-square tests and want a measure of effect size. I have some 1x2 tables, 1x4 tables, 4x9, and 3x4 tables of frequency data.

What measure of effect size can I use on these rectangular frequency tables, if each cell has a different expected frequency?

(I would like to use something like Cramer's V, but it only applies to tables with the same expected frequencies.)

Sample data: two groups, 'false' observes 535 (expected: 3/4), 'true' observes 222 (expected: 1/4).

I think you're wrong about Cramer's $V$ requiring equal expected frequencies...Do you have a source to support your claim? I can't find any myself, and the formula certainly doesn't require equal frequencies to produce a result. — Nick Stauner, Jan 29, 2014 at 9:43

rolando2 · Accepted Answer · 2011-07-14 01:32:26Z

4

Actually Cramer's V will work for non-square tables. See Measures of Association or Wikipedia

answered Jul 14, 2011 at 1:32

rolando2

12.5k1 gold badge42 silver badges63 bronze badges

$\begingroup$ My fault. I meant that it does not work for non-square tables with different expected frequencies. I have edited the original question. $\endgroup$
– newresearcher
Jul 14, 2011 at 14:27

Add a comment |

1 Answer 1